PFAS

Ejemplo PFAS

Primero importamos las librerias ncesesarias.

import pandas as pd
import numpy as np
import plotly.express as px

pfas_data = pd.read_csv("../data/pfas_data.csv")

pfas_data.head()

pfas_data.info()

pfas_data["RDKIT_SMILES"].isna().mean()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 6071 entries, 0 to 6070
Columns: 2091 entries, RDKIT_SMILES to PubchemFP880
dtypes: float64(1176), int64(914), object(1)
memory usage: 96.9+ MB

np.float64(0.0)

molecules = pfas_data["RDKIT_SMILES"].tolist()
molecules[:5]

['FC(F)Cl',
 'FC(F)=C(F)F',
 'FC(F)(Cl)C(F)(Cl)Cl',
 'C=C(F)F',
 'OC(C(F)(F)F)C(F)(F)F']

!pip install rdkit

Requirement already satisfied: rdkit in /opt/venv/lib/python3.13/site-packages (2025.3.5)
Requirement already satisfied: numpy in /opt/venv/lib/python3.13/site-packages (from rdkit) (2.3.2)
Requirement already satisfied: Pillow in /opt/venv/lib/python3.13/site-packages (from rdkit) (11.3.0)

from rdkit import Chem
from rdkit.Chem import AllChem

mols = [Chem.MolFromSmiles(smiles) for smiles in molecules]

mols[:5]

[<rdkit.Chem.rdchem.Mol at 0xffff701bdaf0>,
 <rdkit.Chem.rdchem.Mol at 0xffff701bdc40>,
 <rdkit.Chem.rdchem.Mol at 0xffff701bdcb0>,
 <rdkit.Chem.rdchem.Mol at 0xffff701bdd20>,
 <rdkit.Chem.rdchem.Mol at 0xffff701bdd90>]

fps = [AllChem.GetMACCSKeysFingerprint(mol) for mol in mols]

fps[:5]

[<rdkit.DataStructs.cDataStructs.ExplicitBitVect at 0xffff703312a0>,
 <rdkit.DataStructs.cDataStructs.ExplicitBitVect at 0xffff70331620>,
 <rdkit.DataStructs.cDataStructs.ExplicitBitVect at 0xffff70331690>,
 <rdkit.DataStructs.cDataStructs.ExplicitBitVect at 0xffff70331700>,
 <rdkit.DataStructs.cDataStructs.ExplicitBitVect at 0xffff70331770>]

fps_array = np.array(fps)
fps_array

array([[0, 0, 0, ..., 0, 0, 0],
       [0, 0, 0, ..., 0, 0, 0],
       [0, 0, 0, ..., 0, 0, 0],
       ...,
       [0, 0, 0, ..., 1, 0, 0],
       [0, 0, 0, ..., 1, 0, 0],
       [0, 0, 0, ..., 1, 0, 0]], shape=(6071, 167))

from sklearn.manifold import TSNE

tsne = TSNE(n_components = 2, perplexity = 50, random_state = 42)

fps_tsne = tsne.fit_transform(fps_array)

tsne_df = pd.DataFrame(data = fps_tsne, columns = ["Component_1", "Component_2"], index = pfas_data["RDKIT_SMILES"])
tsne_df.head()

	Component_1	Component_2
RDKIT_SMILES
FC(F)Cl	-56.922298	60.720055
FC(F)=C(F)F	-69.412216	26.313509
FC(F)(Cl)C(F)(Cl)Cl	-32.573349	29.395290
C=C(F)F	-69.444672	26.623600
OC(C(F)(F)F)C(F)(F)F	-22.384430	-27.337769

pfas_classes = pd.read_csv("../data/pfas_classes.csv", index_col = "RDKIT_SMILES")

tsne_df_joined = pd.merge(tsne_df, pfas_classes, on = "RDKIT_SMILES", how = "inner")
tsne_df_joined.head()

	Component_1	Component_2	First_Class	Second_Class
RDKIT_SMILES
FC(F)Cl	-56.922298	60.720055	PFAS derivatives	PFAS halogen derivatives
FC(F)=C(F)F	-69.412216	26.313509	PFAS derivatives	With fluorinated C=C or C=O carbon
FC(F)(Cl)C(F)(Cl)Cl	-32.573349	29.395290	PFAS derivatives	PFAS halogen derivatives
C=C(F)F	-69.444672	26.623600	PFAS derivatives	With fluorinated C=C or C=O carbon
OC(C(F)(F)F)C(F)(F)F	-22.384430	-27.337769	Other aliphatics	Others

fig = px.scatter(tsne_df_joined,
                 x = "Component_1",
                 y = "Component_2",
                 color = "First_Class",
                 hover_name = tsne_df_joined.index,
                 hover_data = {"Component_1": ":.2f",
                               "Component_2": ":.2f",
                               "First_Class": True})

fig.update_layout(title = "Moléculas de PFAS mediante t-SNE",
                  xaxis_title = "Componente t-SNE 1",
                  yaxis_title = "Componente t-SNE 2",
                  height = 600, width = 900)
fig.show();

Ahora vamos a hacerlo en R3

tsne = TSNE(n_components = 3, perplexity = 50, random_state = 42)

fps_tsne = tsne.fit_transform(fps_array)

tsne_df = pd.DataFrame(data = fps_tsne,
                     columns = ["Component_1", "Component_2", "Component_3"],
                     index = pfas_data["RDKIT_SMILES"])

tsne_df_joined = pd.merge(tsne_df, pfas_classes, on = "RDKIT_SMILES", how = "inner")
tsne_df_joined.head()

	Component_1	Component_2	Component_3	First_Class	Second_Class
RDKIT_SMILES
FC(F)Cl	-15.876364	24.557077	7.744306	PFAS derivatives	PFAS halogen derivatives
FC(F)=C(F)F	-24.740488	3.906794	-2.053560	PFAS derivatives	With fluorinated C=C or C=O carbon
FC(F)(Cl)C(F)(Cl)Cl	-9.861135	1.608409	8.375186	PFAS derivatives	PFAS halogen derivatives
C=C(F)F	-24.910532	3.875210	-1.773332	PFAS derivatives	With fluorinated C=C or C=O carbon
OC(C(F)(F)F)C(F)(F)F	-8.248621	-9.961960	-0.368669	Other aliphatics	Others

fig = px.scatter_3d(tsne_df_joined,
                    x = "Component_1",
                    y = "Component_2",
                    z = "Component_3",
                    color = "First_Class",
                    hover_name = tsne_df_joined.index,
                    hover_data = {"Component_1": ":.2f",
                                  "Component_2": ":.2f",
                                  "Component_3": ":.2f",
                                  "First_Class": True})

fig.update_layout(scene = dict(xaxis_title = "Componente t-SNE 1",
                               yaxis_title = "Componente t-SNE 2",
                               zaxis_title = "Componente t-SNE 3"))
fig.show();